supermdguy

6 ms·

31.
by supermdguy 5mo ago
Interesting to see this after the recent post about Chrome’s on-device model using up 4gb of storage, which frustrated a lot of people [1]. I agree local models are great, and it’s cool that Apple has models built in now. But I feel like it
32.
by supermdguy 5mo ago
Better auth is great! I love how it's way more hackable than using a something like Clerk. We were able to add a plugin to allow auth via iframe postMessage (embedded in a CRM) and everything worked seamlessly.
33.
by supermdguy 5mo ago
I've used it off and on over the last month or so. For more complicated tasks (30+ minutes) it works well, and seems to replace a lot of prompting that I'd normally need to do (e.g. asking questions about requirements, creating sp
34.
by supermdguy 5mo ago
There’s a really good analysis here: https://www.lesswrong.com/posts/LBC2TnHK8cZAimdWF/will-jesus... > The Yes people are betting that, later this year, their counterparties (the No betters) will want cash (to
35.
Ask.com has closed(ask.com)
477 points by supermdguy 5mo ago|240 comments
36.
by supermdguy 6mo ago
Overall, I'm really impressed by what you accomplished! I'm not a researcher, so not sure if this is that helpful, but here are some thoughts: - I wonder if the "move" action is difficult for the model to learn to use we
37.
by supermdguy 6mo ago
Cool concept! I think the hardest part will be getting people in the target audience to use it. A lot of indie hackers make software for other indie hackers, but that isn't true of most other verticals. And honestly building software f
38.
by supermdguy 6mo ago
From the most recent comment, looks like this is a bug, triggered by the system inadvertently activating an internal release tool [0]. Still a pretty wild bug, but not as dramatic as the title suggests. Which is kind of unfortunate honestly
39.
by supermdguy 6mo ago
Next step is to build an analog scientific calculator with only EML gates
40.
by supermdguy 6mo ago
Bizarre reading the thread, it feels like their Claude responding to the other posters’ Claudes
41.
The Hypercurious Mind(aeon.co)
3 points by supermdguy 6mo ago|1 comments
42.
by supermdguy 6mo ago
If all LLM advancements stopped today, but compute + energy got to the price where the $30 million zettaflop was possible, I wonder what outcomes would be possible? Would 1000 claudes be able to coordinate in meaningful ways? How much human
43.
by supermdguy 6mo ago
And also OpenAI’s codex spark?
44.
by supermdguy 6mo ago
Headline/article is extremely misleading. They still have subscription plans with included usage, but those usage limits are now based on tokens instead of messages. https://help.openai.com/en/articles/2000110
45.
by supermdguy 6mo ago
I like this, and think it's true for how humans learn. What's interesting to me is that it seems LLMs are significantly smarter than they were two years ago, but it doesn't feel like they have better "taste". Their
46.
by supermdguy 6mo ago
Correct URL: https://www.gingerbill.org/article/2026/02/21/does-syntax-ma...
47.
by supermdguy 6mo ago
I’ve tried having one “big” task that I’m focusing on with active back and forth while letting other Claude instances handle easier back-burner type tasks that it can effectively one-shot. But I’ve noticed that often turns into me spending
48.
by supermdguy 6mo ago
Yeah that confused me, but the compression paper also doesn’t make a ton of sense since I doubt Google would have released it if it was actually such a competitive advantage compared to what other labs are doing. So I wonder what’s actually
49.
by supermdguy 6mo ago
Any other sources on the OpenAI claim? Regardless, it’ll be nice to have cheaper RAM
50.
by supermdguy 6mo ago
Okay this is really fun and mathematically satisfying. Could even be useful for tough bugs that are technically deterministic, but you might not have precise reproduction steps. Does it support running a test multiple times to get a probabi
51.
You can't imitation-learn how to continual-learn(lesswrong.com)
11 points by supermdguy 6mo ago|0 comments
52.
UK startup ignites plasma inside nuclear fusion rocket(euronews.com)
3 points by supermdguy 6mo ago|0 comments
53.
Dionne Quintuplets(en.wikipedia.org)
1 points by supermdguy 6mo ago|0 comments
54.
by supermdguy 6mo ago
It's surprising that this works so well considering that AI-generated AGENTS.md files have been shown to be not very useful. I think the key difference here is that the real-world experience helps the agent reach regions of its latent
55.
by supermdguy 6mo ago
> Nobody knows yet the true capabilities of the missile, but it doesn’t matter. The accuracy doesn’t matter very much, the payload doesn’t matter very much. If it’s launched at a certain target in Tel Aviv, it still is going to hit somet
56.
by supermdguy 7mo ago
> Experimenting with ideas/refactors to see how they'll play out (often the agent can just tell you how it's going to play out) This has helped me a lot. Normally I'd feel really attached to big refactors because of s
57.
by supermdguy 7mo ago
> One might note that MCTS uses more inference compute on a per-sample basis than GRPO: of course it performs better This part confused me, it sounded like they were only doing the MCTS at train time, and then using GRPO to distill the M
58.
by supermdguy 7mo ago
I’ve actually written a crawler like that before, and still ended up going with Firecrawl for a more recent project. There’s just so many headaches at scale: OOMs from heavy pages, proxies for sites that block cloud IPs, handling nested ifr
59.
by supermdguy 7mo ago
I’ve been trying to learn a lot about domain driven design, I think knowledge crunching will be a huge part of the new software development role.
60.
by supermdguy 7mo ago
I'd also add errors with third-party systems, which aren't the developer's or the user's fault, but which are probably worth handling nicely (e.g. retry with backoff).

More ›